4165 Offseason swerve

[matthewdenny]

Here is the plan for our offseason swerve (I've attached the ppt, explaining how we are making it, and the advantages and specs)

Whats different with this swerve? There is basically no machining involved (1 part needs a shaft turned down on a lathe slightly)

All drive power and steering are all gear transfer (no belts or chains).

Revletively compact (5"x5"x10")

~7.4 lbs

(there is an un-pictured thrust bearing above the bevel box)

Looking for feedback as we order parts to build it.

Also a big thanks to team 2590 for an Innovation grant that will help us afford to do this.

[String]

If you want to get rid of lathe use, for teams that have a drill press (probably more common than a lathe), this would be a potential modification. Use the standard VexPro gear ratios in the gearbox, and use 2x2 (or bigger) aluminum tube extrusion to house bevel gears. Using a big enough bevel gear to broach a 1/2" hex (http://www.mcmaster.com/#7297k18/=wnul54), teams could drill bearing holes themselves and use nothing but 1/2" hex throughout the entire swerve module. I may be missing something, I doubt I've done as much research as you guys, and I know this takes away some of that COTS/simplicity of assembly aspect you were going for. It might still be worth looking into for teams without a lathe regardless.
In essence, you can do some interesting things if you let go of the bevel box.

[thedude019]

So what are the advantages of using two wheels instead of one single wheel? I could imagine there being more friction when trying to turn the wheels in place. Or just turning in general.

[Ether]

Can you show your calculations for the thrust, side, and cantilever loading of the bevel gear bearing?

[matthewdenny]

The friction isn't really an issue because the each wheel now only supports half as much weight. The advantage for us is that there is less torque on the shaft supporting the wheel going through the bevel box and on the associated bearings.

It may ultimately end up being okay to have only 1 wheel, but we'd have to test to see. Also with one wheel depending on your maneuvering the actual wheel base would constantly be changing slightly in size, which could affect how you turn. This could probably be addressed in code if needed though by creating a dynamic center point of the robot.

Also we thought it looked kinda cool.

[Ether]

Originally Posted by matthewdenny

The friction isn't really an issue because the each wheel now only supports half as much weight.

I think Luke was comparing your 2-wheel arrangement to a swerve with only one wheel centered on the turning axis. In that comparison, it takes much more torque to steer the 2-wheel arrangement. Also, the kinematics with the 2-wheel arrangement are slightly compromised: there will be some wheel scrubbing.

Have you done calculations on the loading of that bevel gear bearing? It will see thrust, side, and cantilever loading.

[matthewdenny]

Originally Posted by Ether

I think Luke was comparing your 2-wheel arrangement to a swerve with only one wheel centered on the turning axis. In that comparison, it takes much more torque to steer the 2-wheel arrangement. Also, the kinematics with the 2-wheel arrangement are slightly compromised: there will be some wheel scrubbing.

Have you done calculations on the loading of that bevel gear bearing? It will see thrust, side, and cantilever loading.

Ok, got it. Yes there is noticeably more friction with the wheels out from the turning axis, but we calculate that we still can turn 90deg with a fully weighted robot in <.1s, which was acceptable for us. The 2 wheel arrangement allowed us to lower the cg, and keep everything COTS, which was a primary design concern.

As far as the bearing goes, we aren't sure where to find the spec on the bearing load limits or what kind of shock forces we can expect. We do anticipate that most thrust forces would be held by a needle roller bearing between the bevel box upper gear and vex gearbox though. The side and cantelever loads are a bit of a mystery, and any insight you could share on calculating that would be appreciated.

[matthewdenny]

I would appreciate any info you can give on calculating the load on the bearing however.

[Joe Johnson]

Originally Posted by matthewdenny

Ok, got it. Yes there is noticeably more friction with the wheels out from the turning axis, but we calculate that we still can turn 90deg with a fully weighted robot in <.1s, which was acceptable for us. The 2 wheel arrangement allowed us to lower the cg, and keep everything COTS, which was a primary design concern.

As far as the bearing goes, we aren't sure where to find the spec on the bearing load limits or what kind of shock forces we can expect. We do anticipate that most thrust forces would be held by a needle roller bearing between the bevel box upper gear and vex gearbox though. The side and cantelever loads are a bit of a mystery, and any insight you could share on calculating that would be appreciated.

I fear your steering motors will be very unhappy.

You have a 48:1 ratio from the BB to the swerve axle. Let's say that each stage is 90% efficient, two stages, that is an effective ratio of 39:1. The BB has a stall torque at 12V of 0.8N-m so you've got 31N-m at the swerve axles.

BUT, your wheels look to be perhaps 4-5" on center, let's call it 110mm and be done, that means you're 31N-m of max steering torque can put about 280N as its dragging those grippy wheels over the carpet. That's about 60lbs. No problem you think to yourself. Even if the robot full weight in on one swerve module and the CoF of the wheels is 1.0, your BB motors can do the job. And you'd be right, they can turn the wheels, bearly. But they are not going to be happy about doing it for long periods of time.

Let's do the less drastic case: 120lbs robot, 1/4 on each wheel. That is a normal force per wheel of 15 lbs (67N). With a CoF of 1.0, it takes over 7N-m just to drag the tires over the carpet (which you have to do if you want to turn the wheels) 7N-m is about 1/4 of the stall torque you have available and will take 20A for each swerve module. Yes, the BB can to do this but again, not for long periods of time.

I am happy that you are trying this, I just want to put some things out the for you to consider.

Dr. Joe J.

[matthewdenny]

Is it a heat issue with running 20A through the motors? Do you think that normal match play, with periodic travel in strait lines and pauses periodically for 2.5 mins would be a problem?

Originally Posted by Joe Johnson

I fear your steering motors will be very unhappy.

You have a 48:1 ratio from the BB to the swerve axle. Let's say that each stage is 90% efficient, two stages, that is an effective ratio of 39:1. The BB has a stall torque at 12V of 0.8N-m so you've got 31N-m at the swerve axles.

BUT, your wheels look to be perhaps 4-5" on center, let's call it 110mm and be done, that means you're 31N-m of max steering torque can put about 280N as its dragging those grippy wheels over the carpet. That's about 60lbs. No problem you think to yourself. Even if the robot full weight in on one swerve module and the CoF of the wheels is 1.0, your BB motors can do the job. And you'd be right, they can turn the wheels, bearly. But they are not going to be happy about doing it for long periods of time.

Let's do the less drastic case: 120lbs robot, 1/4 on each wheel. That is a normal force per wheel of 15 lbs (67N). With a CoF of 1.0, it takes over 7N-m just to drag the tires over the carpet (which you have to do if you want to turn the wheels) 7N-m is about 1/4 of the stall torque you have available and will take 20A for each swerve module. Yes, the BB can to do this but again, not for long periods of time.

I am happy that you are trying this, I just want to put some things out the for you to consider.

Dr. Joe J.

[cxcad]

Originally Posted by matthewdenny

Is it a heat issue with running 20A through the motors? Do you think that normal match play, with periodic travel in strait lines and pauses periodically for 2.5 mins would be a problem?

Yes, temperature of the motors is proportional to current squared. You will also be turning your modules more often then you think you would due to software controls and subconscious manual fine adjustments

[matthewdenny]

Heres my back of the napkin calculations:

the motor is using ~23 Amps @ 12 V for a power of 276 W

The banebots page lists the motor at 78% efficient, so I'm going to round that down to 65%

so 35% of the 276 W goes to heat = 96 W of heat

96W * 150 seconds of game time* 70% duty cycle during match = 10100 Joules of heat

Q=mC(deltaT) ... m=337 g Csteel= .47

deltaT = 64 C

Which would make it pretty hot, but the motor blows air as it is moving, and conducts heat to the gearbox and surrounding air. Not sure how to calculate the convection through it.

[Kevin Ainsworth]

We run 100:1 on the steering gearbox with 2:1 from the gearbox to the steering wheel.
So 200:1 total reduction using a BB 550 motor.
And we only use one 1" wide wheel.
You also need to think about what happens in a defensive year when the bot gets loaded to one side and the wheel is trying to steer.

[Bryce2471]

Originally Posted by Kevin Ainsworth

We run 100:1 on the steering gearbox with 2:1 from the gearbox to the steering wheel.
So 200:1 total reduction using a BB 550 motor.
And we only use one 1" wide wheel.
You also need to think about what happens in a defensive year when the bot gets loaded to one side and the wheel is trying to steer.

200:1 seems a little conservative, but it is better to error on the side of robustness when it comes to swerve gearing.

I think we have hit the sweet spot this year between cool motors, and responsive controls. We are using a 64:1 P60 gearbox, fallowed by a 2:1 belt reduction for ~130:1 final reduction, on a 550 motor.

Last year, we had an 86:1 final reduction on an AM 9015. We never lost a motor because we used a torque limiting gearbox to prevent stall conditions. But the motors would frequently get hot, and the performance in pushing matches was lackluster.